Build evaluation criteria for a procurement comparison
Translate business needs, risks, and constraints into observable vendor evaluation criteria before reviewing offers
10 min setupTested with:ChatGPTReviewed: 2026-08-28
1
Add your context
Your text stays in this browser. AILesson Prompts does not send it to a model or server.
2
Your prompt
Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt
Build a fair procurement evaluation framework before vendor scoring.
Users, outcome, current state, demand, locations, duration, budget, schedule, dependencies, and exclusions:
[need]
Functional, service, security, privacy, compliance, accessibility, implementation, support, continuity, sustainability, commercial, and exit needs:
[requirements]
Stakeholders, approvers, rules, conflicts, scale, weighting authority, evidence, due diligence, and decision date:
[governance]
Separate mandatory gates, scored criteria, preferences, contract terms, and due-diligence checks. Map every criterion to an approved need or risk; remove duplicates and vendor-specific wording. Define observable evidence, measurement unit, applicable scenario, rating anchors for every score, missing-evidence treatment, and owner. Use total cost of ownership rather than headline price where inputs permit, without inventing volumes, taxes, exchange rates, discounts, implementation, renewal, termination, or exit cost. Set weights and pass thresholds only when authorized, verify arithmetic, and test sensitivity to plausible weight changes. Prevent double-counting related qualities and do not use reputation, incumbent status, presentation polish, relationship, or unsupported sustainability claims as proxies. Include accessibility, privacy, security, continuity, implementation, service, commercial, and exit risks where relevant. End with conflicts declaration, independent scoring sequence, clarification rules, calibration cases, decision record, and required legal/security/finance/procurement reviews.
Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.
From input to outcome
A worked example
See how concrete context turns this recipe into a usable result
Actual input
Business need and scope
Select a help-desk platform for 38 support agents in UK and Germany for a 3-year term. Must replace current contract by Jan 31 2027. Required outcomes: manage email and chat cases, preserve audit trail, provide English and German agent UI, export all ticket data on exit. Budget ceiling GBP 95,000 over three years including implementation and mandatory fees. Existing ticket volume 18,000/year. No need for voice channel.
Requirements and risks
Mandatory: EU or UK data hosting approved by Security, SSO/SAML, role-based access, deletion workflow, WCAG 2.2 AA evidence for agent interface, email/chat, German UI, complete ticket export. Desired: implementation by Jan 10, sandbox, support response SLA, documented uptime, admin training. Commercial: all mandatory fees, annual increases, implementation, training, support, data export and termination. Security and legal due diligence required. Sustainability evidence is a preference but no approved metric.
Email and chat case management with audit trail, English/German agent UI: scripted scenario; Support. Voice is excluded and must not earn points.
WCAG 2.2 AA evidence for the agent interface: current conformance report plus targeted validation; Accessibility. A vendor assertion alone is insufficient.
Complete ticket export on exit: schema/sample export and contract right; Procurement/Legal.
Three-year mandatory TCO no more than GBP 95,000: Finance calculation using disclosed fees. Missing gate evidence may be clarified only by the same fixed deadline for all vendors.
Weighted criteria, total 100
Functional fit 25; implementation 15; service/continuity 15; accessibility quality beyond the gate 10; security/privacy quality beyond the gate 15; three-year TCO 20. To avoid double-counting, gate compliance itself earns no scored points; scored security/accessibility evaluates evidence depth and risk above minimum. Sustainability remains an unscored preference until an approved measure exists.
Common 0-4 anchors: 0 no usable evidence or fails scenario; 1 major gaps/high reliance on unconfirmed work; 2 meets basic need with material limitations; 3 meets need with credible evidence and manageable risk; 4 exceeds the relevant scenario with independently reviewable evidence and no material added complexity. Each criterion needs scenario-specific anchor detail before issue.
TCO formula: three years of license for 38 agents + mandatory implementation + training + required support + known annual increases + mandatory export/termination charges + applicable taxes, which remain unknown. Do not score discounts unless contractually offered.
Process: conflict declarations; gate review; independent criterion scoring with evidence locators; clarification; calibrated panel discussion; sensitivity test at ±5 weight points moved between TCO and functional fit while total stays 100; shortlist; security/legal due diligence; CFO approval. Record whether ranking changes under sensitivity and all unresolved contract risks.
Why this works
1
Predeclared criteria reduce the chance that a persuasive proposal changes the rules after evidence is seen.
2
Gates and scored trade-offs separate unacceptable risk from compensable weakness.
Check the result
Does every gate and criterion trace to an approved need or risk?
Are evidence, units, anchors, missing-data rules, and owners explicit?
Are weights authorized, arithmetically correct, non-duplicative, and sensitivity-tested?
ChatGPT